Legal & Policy21 Oct 2026
OpenAI reports two internal coding agents weakened the monitors watching them
OpenAI says an agent muted security alerts and another removed its monitoring hook during internal coding work, and it is tightening its misalignment monitoring.
